77 Predicting the tumor microenvironment molecular composition to assess immunotherapy efficacy in non-small cell lung cancer from digital histopathology images
Year: 2024
Type: conference-abstract
Abstract: Background Immune checkpoint inhibitors (ICI) have become integral to non-small cell lung cancer (NSCLC) treatment. However, biomarkers predictive of immunotherapy efficacy are limited. Programmed Death Ligand 1 (PD-L1) expression has been clinically approved to identify responders to ICI, but despite selecting for high PD-L1 expression, response rates still vary widely, ranging from 17-49%.1 Hence, there is a clinically unmet need for additional predictive biomarkers capturing tumor and microenvironmental factors associated with ICI response. With artificial intelligence, there is great potential to unlock TME-related information related to Hematoxylin and Eosin (H&E)-stained pathology slides, which are readily accessible through clinical routine.2 Methods We developed HistoTME, a novel weakly supervised deep learning approach to infer the tumor microenvironment (TME) molecular composition directly from histopathology images of NSCLC patients. The model learns to predict the gene expression levels of 30 cell type-specific TME signatures from H&E whole slide images (figure 1). HistoTME was trained using H&E images and matched bulk transcriptomics data from the TCGA-NSCLC cohort (N=865 patients) and validated using an external cohort of CPTAC-NSCLC (N=333 patients). Next, TME signatures were used to characterize TME status (inflamed vs desert) and to predict response to ICI. These methods were tested retrospectively using survival outcomes in an external cohort of NSCLC patients from SUNY Upstate (N=292 patients) who were treated with immune checkpoint inhibitor therapy. Results HistoTME accurately predicts the expression of 30 TME molecular signatures from H&E images, achieving an average Pearson correlation of 0.50 with the ground truth. With HistoTME-predicted signatures, we derived TME status (inflamed vs desert), improving prognostication of patients receiving first-line ICI treatment (figure 2A, p-value=0.0012, HR = 0.52[95% CI: 0.35-0.77]). Patients with PDL1 expression ≥50% also showed improved survival (figure 2B, p-value=0.0059, HR=0.62 [95% CI: 0.37- 1.04]). In additional subgroup analysis, we find that the H&E-inferred TME status is predictive of overall survival outcomes of PD-L1 absent (<1%) patients (figure 2C; p-value=0.08, HR=0.4[95% CI: 0.13- 1.18]), and PD-L1 low (1-49%) patients (figure 2D; p-value=0.009, HR=0.44[95% CI: 0.23-0.82]) but not PD-L1 high (≥50%) patients (figure 2E; p-value=0.85, HR=0.94[95% CI: 0.48-1.84]). Additionally, the signatures could be utilized to accurately predict response, achieving an AUROC of 0.75[95% CI: 0.61-0.88] for predicting responses following first-line ICI treatment. Conclusions Collectively, HistoTME presents an effective approach for characterizing the TME molecular composition and predicting ICI response for NSCLC patients, complementing the current clinical biomarker, PD-L1 expression, and bringing us closer to personalized immuno-oncology. References Sholl LM, Awad M, Basu Roy U, Beasley MB, Cartun RW, Hwang DM, et al. Programmed death ligand-1 and tumor mutation burden testing of patients with lung cancer for selection of immune checkpoint inhibitor therapies: guideline from the college of american pathologists, association for molecular pathology, international association for the study of lung cancer, pulmonary pathology society, and LUNGevity foundation. Arch Pathol Lab Med 2024. Wang X, Barrera C, Bera K, Viswanathan VS, Azarianpour-Esfahani S, Koyuncu C, et al. Spatial interplay patterns of cancer nuclei and tumor-infiltrating lymphocytes (TILs) predict clinical benefit for immune checkpoint inhibitors. Sci Adv 2022;8(22):eabn3966. Ethics Approval The study was reviewed by the ethics committee at SUNY Upstate and considered exempt from IRB oversight.
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4404064986